By Kumar Navulur, Bill Baugh, Greg Hammann and Vic Leonard, DigitalGlobe (www.digitalglobe.com), Longmont, Colo.
After more than 40 years, the remote sensing community continues to face two fundamental challenges when using Earth imagery at a global scale: automated information extraction and change detection. DigitalGlobe’s WorldView-3 satellite, currently being developed by Ball Aerospace for a 2014 launch, is designed to address these challenges by creating consistent datasets as well as providing unique information for agriculture, forestry and mining/geology applications (see “Exploring the Benefits of SWIR Satellite Imagery,” below).
Doubling the Spectral Bands
WorldView-3 will be the first satellite to have 16 high-resolution spectral bands that capture information in the visible and near-infrared (VNIR) and short-wave infrared (SWIR) regions of the electromagnetic spectrum (EMS). Operating at an expected altitude of 617 kilometers, the satellite will provide 31-centimeter panchromatic resolution, 1.24-meter VNIR resolution and 3.7-meter SWIR resolution.
WorldView-3 builds upon WorldView-2’s VNIR capabilities, providing eight additional spectral bands farther into the SWIR portion of the EMS. This spectral resolution expansion enhances WorldView-3’s capability to capture the uniqueness of each ground material’s spectral signature. Due to minimal atmospheric influence or noise in this part of the EMS, as well as an enhanced ability to differentiate among ground materials, the SWIR bands open the door for automated information extraction to save time, money and possibly lives.
Creating Consistent Imagery
Remote sensing satellites view Earth from above the atmosphere-—top-of-the-atmosphere (TOA) measurements. However, materials on Earth’s surface absorb and reflect light at surface level, so only surface-reflectance data truly represent the nature of the materials.
Changes in the atmosphere, sun illumination and viewing geometries during image capture result in inconsistent image data, hindering automated information extraction and change detection. Atmospheric conditions typically change during and between different imagery collections due to varying moisture levels (water vapor) and particulates (aerosols) in the atmosphere. Much research has been done trying to accurately convert the TOA measurements to surface-reflectance measurements.
Several models have been developed to compensate for these atmospheric conditions and the viewing geometries of various satellites. In addition, the remote sensing research community has created normalized indices to counter some atmospheric conditions with limited success.
The challenge has been the availability of accurate atmospheric measurements at appropriate scale, ensuring imagery can be normalized. WorldView-3 is expected to address this problem by becoming the first commercial imaging satellite with an atmospheric sensor as part of its payload.
During image capture, the WorldView-3 atmospheric sensor is designed to detect the presence of clouds, aerosols and water vapor at 30-meter resolution, thereby measuring the exact atmospheric conditions corresponding to every recorded image. Figure 1 shows how the atmospheric sensor has a slightly wider swath than the imaging swath.
DigitalGlobe has developed proprietary algorithms that use these atmospheric measurements to normalize WorldView-3 imagery for consistency. This normalization is called atmospheric compensation, which is especially important for information extraction, such as change detection and vegetation analysis, because changes due to the atmosphere have been removed. Atmospheric compensation results in surface-reflectance image data. Figure 2 shows an example of a surface-reflectance image after atmospheric compensation. The Normalized Difference Vegetation Index from data without atmospheric compensation underestimates the amount of vegetation by about 10-13 percent.
Another issue impairing automated information extraction is accurately mapping cloud cover. WorldView-3’s sensors have spectral bands that range from the VNIR into the SWIR part of the EMS to accurately distinguish clouds from other bright features such as snow and ice.
Figure 3 shows how the longer wavelengths in the SWIR range of the atmospheric sensor are able to penetrate fire smoke and haze, allowing for clouds to be more accurately delineated. Figure 4 shows an example of a 2010 volcano in Iceland where WorldView-3’s super-spectral sensors will be able to differentiate between ash, ice and clouds.
WorldView-3 will be the first super-spectral satellite to simultaneously map atmospheric conditions during image collection, allowing unprecedented access to normalized imagery across the globe. Such standardization will introduce a new age in automated information extraction and change detection.
Moving from Pixels to Products
WorldView-3’s atmospheric sensor will be used to normalize imagery for varying atmospheric conditions and to develop algorithms that can be used anywhere on the globe. Furthermore, the satellite’s 16 spectral bands will allow for automated information extraction for various applications. Because WorldView-3 was designed as an evolutionary and revolutionary sensor, the satellite will transform the remote sensing industry from a pixel-based industry into a product-based industry, expanding the use of remotely sensed data to create ways to better understand and manage our changing planet.